Prognostic prediction model of colorectal cancer based on preoperative serum tumor markers

Abstract

BACKGROUND

Preoperative serum tumor markers not only play a role in the auxiliary diagnosis and postoperative monitoring in colorectal cancer (CRC), but also have been found to have potential prognostic value.

AIM

To analyze whether preoperative serum tumor markers, including carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9), affect the prognosis of CRC.

METHODS

This was a retrospective study conducted in a single center. Patients with nonmetastatic CRC who underwent initial surgery between January 2011 and January 2020 were enrolled and divided into development site and validation site groups at a ratio of 7:3. The independent prognostic factors were screened by Cox regression analysis, and finally, a prognostic nomogram model was established. The newly developed model was tested by internal validation.

RESULTS

Eventually, 3526 postoperative patients with nonmetastatic CRC were included in the study. There were 2473 patients at the development site and 1056 patients at the validation site. Age (P < 0.01, HR = 1.042, 95%CI = 1.033-1.051), tumor node metastasis (TNM) classification (P < 0.01, HR = 1.938, 95%CI = 1.665-2.255), preoperative CEA (P = 0.001, HR = 1.393, 95%CI = 1.137-1.707) and CA19-9 (P < 0.01, HR = 1.948, 95%CI = 1.614-2.438) levels were considered independent prognostic factors for patients with nonmetastatic CRC and were used as variables in the nomogram model. The areas under the curve of the development and validation sites were 0.655 and 0.658, respectively. The calibration plot also showed the significant performance of the newly established nomogram.

CONCLUSION

We successfully constructed a nomogram model based on age, TNM stage, preoperative CEA, and CA19-9 levels to evaluate the overall survival of patients with nonmetastatic CRC.

Key Words: Colorectal cancer; Prognosis; Carcinoembryonic antigen; Carbohydrate antigen 19-9; Nomogram

Core Tip: The tumor markers carcinoembryonic antigen and carbohydrate antigen 19-9 are often used in the auxiliary diagnosis and postoperative monitoring of colorectal patients, but their role in prognosis needs to be further explored. Here, we retrospectively analyzed the clinical data and pathological characteristics of 3526 patients with nonmetastatic colorectal cancer (CRC), confirmed the negative correlation between preoperative serum tumor marker levels and prognosis, and established a nomogram model to evaluate the prognosis of CRC patients.

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer worldwide[1,2]. Although the 5-year survival rate of CRC patients has improved due to continuous improvements in screening, chemoradiotherapy, immunotherapy, metastasis resection and other treatment measures[3,4], CRC is still the second leading cause of cancer death worldwide, accounting for 10% of all cancer deaths[5,6].

At present, the prognosis prediction and treatment decision of patients with CRC depend on traditional tumor node metastasis (TNM) staging (according to the degree of tumor invasion, lymph node status, and distant metastasis status)[7]. However, as the clinical outcomes of patients with the same stage of CRC vary greatly, it is often inaccurate to judge the prognosis by TNM staging alone, especially for patients with nonmetastatic CRC[8,9]. Therefore, it is necessary to identify biomarkers to help judge the prognosis of patients with CRC.

As serum tumor markers, carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9) are often used for postoperative monitoring and auxiliary diagnosis of CRC. At present, CEA is the most commonly used biomarker for CRC. In addition to preoperative monitoring, monitoring every 3-6 months after surgery is recommended by guidelines[10,11]. Although there is some controversy regarding the clinical benefit of CA19-9, it is still considered useful for monitoring disease progression in CRC patients without elevated CEA[12]. Many studies have shown that the preoperative CEA level can help to predict the prognosis of patients with CRC[13-15], and the CA19-9 level is also considered to play a role in monitoring recurrence and evaluating prognosis[16-18].

Due to the existence of tumor heterogeneity, it is impractical to predict the prognosis of patients with CRC with a single factor. Therefore, this study combined tumor markers with TNM stage and clinical characteristics of patients to construct a prediction model for the prognosis of CRC patients using a nomogram in hopes of elucidating further prognostic insights.

MATERIALS AND METHODS

Enrollment of patients

In this retrospective study, we enrolled 3529 patients with nonmetastatic CRC who underwent surgical resection at the Department of Gastrointestinal Surgery, the First Affiliated Hospital of Chongqing Medical University, from January 2011 to January 2020. The inclusion criteria were as follows: (1) Aged ≥ 18 years; (2) diagnosed with primary CRC for the first time; (3) underwent radical surgery; and (4) had TNM stage I-III disease. The exclusion criteria were as follows: (1) Previous diagnosis of any malignant tumor; (2) distant metastasis at initial diagnosis of CRC; and (3) lack of clinical parameters and laboratory results. The study was approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University, and informed consent was obtained from the patients (2022-135-2).

Clinical variables selected for analysis

Clinical and pathological variables were obtained from 3526 selected patients, including age, sex, body mass index (BMI), smoking history, drinking history, hypertension status, type 2 diabetes mellitus (T2DM) status, chronic heart disease status, tumor location, tumor size, TNM stage, surgical time, and the levels of the preoperative serum tumor markers CEA, CA19-9, and alpha-fetoprotein. Serum tumor markers were measured within one week before surgery, and normal and abnormal levels were distinguished based on the test results.

Statistical analyses

The independent sample t test was used to analyze the differences between groups for continuous variables, and the chi-square test or Fisher's exact test was used to analyze the differences between groups for categorical variables. Univariate analysis was used to test the relationships between various prognostic predictors and overall survival (OS). Variables with P values less than 0.05 in univariate analysis were used for Cox proportional hazard analysis. The nomogram model was constructed with R software 4.1.2, and then the newly established nomogram was evaluated in the validation site group. The ratio of participants in the development site group to those in the validation site group was 7:3. The areas under the curve (AUCs) and decision curve analysis (DCA)[19] were used to evaluate the performance of the nomogram model, and the accuracy of the model was further evaluated by comparing the predicted results with the actual observation results through the calibration curve[20,21].

RESULTS

Characteristics of the included patients

According to the inclusion criteria, 3529 patients were ultimately enrolled. Patients were randomized at a 7:3 ratio, with 2473 patients assigned to the development site group and 1056 to the validation site group (Table 1). In the development site group, males accounted for 58.4%, females accounted for 41.6%, and the average age was 63.1 years. Moreover, 47.3% of patients had colon cancer, and 60.1% of the patients had a tumor larger than 5 cm. Among them, 649 (25.1%) had hypertension, and 311 (11.5%) had T2DM. According to the TNM classification, 18.9% of patients were in stage I, 39.9% were in stage II, and 35.6% were in stage III. Preoperative CEA and CA19-9 levels were elevated in 36.8% and 20.2% of the patients, respectively.

Table 1 Baseline information between the development and validation cohorts, n (%).

Characteristics	Development (2473)	Validation (1056)	P value
Age, yr	63.1 ± 12.0	61.7 ± 12.3	0.002a
Sex			0.930
Male	1448 (58.4)	620 (59.1)
Female	1025 (41.6)	436 (40.9)
BMI, kg/m2	22.6 ± 3.2	22.8 ± 3.2	0.180
Smoking	942 (37.0)	395 (37.9)	0.700
Drinking	767 (30.8)	310 (29.8)	0.327
Hypertension	649 (25.1)	257 (25.4)	0.235
T2DM	311 (11.5)	123 (11.5)	0.442
CHD	98 (4.3)	52 (4.8)	0.195
Tumor location			0.217
Colon	1166 (47.3)	582 (46.9)
Rectum	1307 (52.7)	474 (53.1)
TNM stage			0.802
I	490 (18.9)	215 (18.4)
II	1067 (39.9)	443 (40.4)
III	916 (35.6)	398 (36.6)
Tumor size			0.705
< 5 cm	1442 (60.1)	623 (60.9)
≥ 5 cm	1031 (39.9)	433 (39.1)
CEA			0.053
Normal	1562 (63.2)	703 (66.6)
Abnormal	911 (36.8)	353 (33.4)
CA199			0.823
Normal	1973 (79.8)	839 (79.5)
Abnormal	500 (20.2)	217 (20.5)
AFP			0.105
Normal	2366 (95.7)	997 (94.4)
Abnormal	107 (4.3)	59 (5.6)
Surgical time, min	224.5 ± 78.3	224.6 ± 78.3	0.967

^aP < 0.05.

T2DM: Type 2 diabetes mellitus; BMI: Body mass index; CHD: Chronic heart disease; TNM: Tumor node metastasis; AFP: Alpha fetoprotein; CEA: Carcinoembryonic antigen; CA199: Carbohydrate antigen 19-9.

Establishment of the nomogram

To predict the prognosis of CRC, univariate and Cox analyses were performed (Table 2). According to univariate analysis, age (P < 0.01, HR = 1.045, 95%CI = 1.037-1.055), BMI (P = 0.001, HR = 0.952, 95%CI = 0.924-0.981), tumor size (P < 0.01, HR = 1.426, 95%CI = 1.184-1.718), tumor stage (P < 0.01, HR = 2.105, 95%CI = 1.817-2.438), and preoperative CEA (P < 0.01, HR = 2.185, 95%CI = 1.812-2.633) and CA19-9 (P < 0.01, HR = 2.646, 95%CI = 2.185-3.204) levels all showed highly significant differences. Unhealthy lifestyle habits, such as smoking (P = 0.706, HR = 0.964, 95%CI = 0.795-1.168) and drinking (P = 0.248, HR = 0.884, 95%CI = 0.718-1.089), were not significantly associated with OS, and chronic diseases, such as hypertension (P = 0.493, HR = 0.926, 95%CI = 0.743-1.154) and T2DM (P = 0.134, HR = 1.231, 95%CI = 0.938-1.617), were also not associated with OS.

Table 2 Univariate and multivariate analysis of overall survival.

Risk factors	Univariate analysis		Multivariate analysis
	HR (95%CI)	P value	HR (95%CI)	P value
Age (yr)	1.045 (1.037-1.055)	< 0.01a	1.042 (1.033-1.051)	< 0.01a
Sex (male/female)	0.931 (0.770-1.126)	0.461
BMI (kg/m2)	0.952 (0.924-0.981)	0.001a	0.980 (0.952-1.008)	0.161
T2DM (yes/no)	1.231 (0.938-1.617)	0.134
Tumor location (colon/ rectum)	1.179 (0.979-1.421)	0.083
Tumor stage (III/II/I)	2.105 (1.817-2.438)	< 0.01a	1.938 (1.665-2.255)	< 0.01a
Smoking (yes/no)	0.964 (0.795-1.168)	0.706
Drinking (yes/no)	0.884 (0.718-1.089)	0.248
Hypertension (yes/no)	0.926 (0.743-1.154)	0.493
Tumor size (≥ 5 cm/< 5 cm)	1.426 (1.184-1.718)	< 0.01a	1.117 (0.923-1.351)	0.255
CEA (abnormal/normal)	2.185 (1.812-2.633)	< 0.01a	1.393 (1.137-1.707)	0.001a
AFP (abnormal/normal)	1.108 (0.715-1.718)	0.645
CA-199 (abnormal/normal)	2.646 (2.185-3.204)	< 0.01a	1.984 (1.614-2.438)	< 0.01a
Surgical time, min	1.001 (1.000-1.002)	0.069

^aP < 0.05.

BMI: Body mass index; T2DM: Type 2 diabetes mellitus; AFP: Alpha fetoprotein; CEA: Carcinoembryonic antigen; CA199: Carbohydrate antigen 19-9.

Next, we included variables with significant differences in the univariate analysis in the Cox analysis, which revealed that age (P < 0.01, HR = 1.042, 95%CI = 1.033-1.051), tumor stage (P < 0.01, HR = 1.938, 95%CI = 1.665-2.255), preoperative CEA (P = 0.001, HR = 1.393, 95%CI = 1.137-1.707) and CA19-9 (P < 0.01, HR = 1.948, 95%CI = 1.614-2.438) levels were independent risk factors for the prognosis of patients with CRC. A nomogram based on the Cox regression model was established (Figure 1). The score of each factor was obtained according to the patient's own condition, and the total score was obtained by adding the four scores. Then, the prognosis of patients with nonmetastatic CRC was estimated according to the total score.

Figure 1 Nomogram for predicting the prognosis of patients with non-metastatic colorectal cancer. CEA: Carcinoembryonic antigen; CA199: Carbohydrate antigen 19-9.

Validation

To verify whether the nomogram was applicable to other datasets, we conducted a validation study using data from 1056 CRC patients at the validation site. Time-dependent receiver operating characteristic curves for the OS-associated nomograms were generated for predicting 1-, 3-, and 5-year survival rates (Figure 2). The AUCs for 5-year survival were 0.655 at the development site and 0.658 at the validation site, which both indicated good predictive ability. The nomogram calibration curve was obtained by comparing the predicted survival rate of the nomogram with the corresponding survival rate obtained by the Kaplan-Meier method (Figure 3). The calibration curves for both the training and validation sets revealed high-quality prediction results. Figure 4 shows the DCA curves, which further confirmed the net benefit of our nomogram model within a certain threshold probability range.

Figure 2 Receiver operating characteristic curves of the nomogram. A: 1-,3- and 5-year receiver operating characteristic (ROC) of nomogram using train set; B: 1-,3- and 5-year ROC of nomogram using validation set. AUC: Area under the curve.

Figure 3 Calibration curves for the nomogram. A: 1-,3- and 5-year calibration curves of overall survival (OS) using training set; B: 1-,3- and 5-year calibration curves of OS using validation set. OS: Overall survival.

Figure 4 Decision curve analysis for the nomograms. A: Decision curve analysis (DCA) of 1-year survival nomogram using training set; B: DCA of 3-year survival nomogram using training set; C: DCA of 3-year survival nomogram using training set; D: DCA of 1-year survival nomogram using validation set; E: DCA of 3-year survival nomogram using validation set; F: DCA of 5-year survival nomogram using validation set.

DISCUSSION

Relying only on the traditional TNM stage to judge the prognosis of patients with CRC[22], especially patients with nonmetastatic CRC, is difficult, and additional influencing factors should be considered[23-25]. Over the past two decades, many molecular biomarkers of CRC have been extensively investigated, but serum tumor markers remain the most commonly used. CEA and CA19-9, which are readily available serum tumor markers, are widely used in the diagnosis and postoperative monitoring of CRC[26,27]. Although previous studies have verified the use of high preoperative CEA and CA19-9 levels as independent predictors of OS and DFS[28-30], few studies have quantified their impact on prognosis.

In this study, we attempted to establish a nomogram including serum tumor markers combined with traditional TNM staging to improve prognosis prediction in patients with CRC. A total of 15 variables were included in the study, and four variables (age, TNM stage, preoperative CEA level and CA19-9 level) were ultimately included in the nomogram. BMI and tumor size were also considered to be associated with prognostic outcome in the univariate analysis, but the association was not strong according to the Cox analysis. This might be because tumor size tends to indicate a greater tumor burden, which is positively correlated with tumor marker levels[31]. For BMI, an inverse association with CEA was considered, possibly due to the hemodilution effect of increased plasma volume observed in patients with high BMIs[32,33]. Other factors, such as poor lifestyle habits and other chronic diseases, did not appear to be associated with prognosis.

Consistent with the results of other studies[34-37], we found that the CEA level was an independent predictor of survival. Compared with normal levels of CEA, elevated preoperative CEA resulted in a 62% increased risk of death[38]. The guidelines also recommend CEA as an effective predictor of OS[39,40]. Notably, previous reports have shown that the significance of postoperative CEA measurements depends on preoperative CEA levels. Almost all patients with high preoperative CEA levels had increased CEA levels at the time of CRC recurrence, but this increase was rarely observed in patients with normal preoperative CEA levels[17,41]. Therefore, we selected the preoperative CEA level as a prognostic predictor. In contrast, previous guidelines did not recommend the use of CA19-9 to assess prognosis[40,42]. However, similar to our findings, several recent studies have also demonstrated the prognostic value of CA19-9[31,43], especially in CRC patients with normal preoperative CEA levels[12,44]. Furthermore, some studies have reported that the combined assessment of preoperative serum CEA and CA19-9 may enhance the diagnostic prediction and prognosis prediction of CRC patients[45].

In our study, we successfully established a novel prognostic model for patients with nonmetastatic CRC. Compared with previous studies, our sample size was quite large, and after internal validation, our prediction model showed good performance. However, the current study has several limitations. First, our study was retrospective and was conducted at a single center, which might have caused selection bias. Second, while we performed internal validation of the prediction model, it would have been better if external validation could have been performed to verify whether our findings were generally applicable.

CONCLUSION

Our study demonstrated the prognostic impact of the tumor markers CEA and CA19-9 and established a more accurate and practical nomogram model for predicting the prognosis of patients with nonmetastatic CRC.

ACKNOWLEDGEMENTS

We acknowledge all the authors whose publications are referred in our article.